Reduced resilience of terrestrial ecosystems locally is not reflected on a global scale

Abstract

Global climate change likely alters the structure and function of vegetation and the stability of terrestrial ecosystems. It is therefore important to assess the factors controlling ecosystem resilience from local to global scales. Here we assess terrestrial vegetation resilience over the past 35 years using early warning indicators calculated from normalized difference vegetation index data. On a local scale we find that climate change reduced the resilience of ecosystems in 64.5% of the global terrestrial vegetated area. Temperature had a greater influence on vegetation resilience than precipitation, while climate mean state had a greater influence than climate variability. However, there is no evidence for decreased ecological resilience on larger scales. Instead, climate warming increased spatial asynchrony of vegetation which buffered the global-scale impacts on resilience. We suggest that the response of terrestrial ecosystem resilience to global climate change is scale-dependent and influenced by spatial asynchrony on the global scale.

Ecosystems are showing symptoms of resilience loss

“…A lower cutoff would enlarge the areas where resilience loss is detected, but also increase the risk of false positives. A similar study using NDVI data estimated up to ~65% of terrestrial ecosystems show early warning of critical transitions, with strong bias towards boreal forest and taiga [39]. The estimates presented here complement previous efforts [30,39] in taking into consideration fixed effects by biome and a pre-processing technique that removes seasonality and long-term variations that can lead to errors, or bias towards high variable environments (higher latitudes).…”

“…These are biomes where the strongest signals of resilience loss were detected with a right shift on the distribution of |∆|. Other studies quantifying terrestrial ecosystems' resilience with NDVI data also show strong signals from tundra and boreal forests [39]. A recent quantification of aridity thresholds shows that up to 28.6% of current dryland area can cross these thresholds by 2100 in the most drastic climate scenarios [40].…”

A resilience sensing system for the biosphere

“…Other studies (table 1) combine statistical indicators, including AR(1) and variance, into a composite indicator [11,12], but generally without clear justification for their relative weighting, and with the drawbacks of e.g. merging aspects of resilience and resistance, and double-counting the effects of recovery rate.…”

“…Some existing studies have started to consider trends in 'resilience' over time, but do not present a mean value as a baseline, and use composite indicators that do not isolate recovery rate [11,12]. Trends in a composite indicator of autocorrelation, standard deviation, skewness and kurtosis of NDVI over 1981-2015 hints at greatest loss of 'resilience' in the tundra [11]. A more confusing composite indicator hints at primary production showing greatest 'resilience' loss in tundra and boreal forest [12].…”

“…Abbreviations: NDVI, normalized difference vegetation index; GPP, gross primary productivity; LAI, leaf area index; VOD, vegetation optical depth; EVI, enhanced vegetation index; SAR, synthetic aperture radar; dNBR, difference normalized burn ratio; DLM , dynamic linear model.scale, locationyearref.vegetation, perturbation index(es) (satellite/ database)time spanresolution (native)‘resilience’ metric(s)recovery rate?space a time global world2014[7]15 land cover typesNDVI (MODIS)2001–20067 km (250 m)16 dAR(1), SD, CV, spectral entropy, spectral scalingyes (AR(1)) world2015[8]AR model fit subsetNDVI (GIMMS)1981–20060.072°1 month (15 d)AR(1)yes world2017[9]all, drought GPP (MODIS)2000–2010(1 km)1 month (1 d)recovery timenot really world2020[10]all, drought, fire LAI (GLASS, GIMMS)1982–2016, 1982–20110.05°, 0.083°8 d, 15 dmax. stress /recovery timeproblematic world2021[11]allNDVI 3 g (GIMMS)1981–20150.083°1 monthACF-1, variance, skewness, kurtosisproblematic (composite indicator)…”